IsValorum/XYZ-Aquila-mini-APEX-I-MiniPlus-V2-GGUF

Quick Navigation Index

Independently computed using dedicated cloud compute clusters, funded 100% out-of-pocket.

If you consider these releases to be of high quality and they help you in your day-to-day workflow, consider supporting on Ko-fi. Every contribution directly sustains future cloud compute hours and helps support my university studies.

Independent Benchmark of the APEX-I-MiniPlus Family (Occamy V2 Reference)
Bundled Model Files & Specifications
Comparative Quantization Analysis (vs. Flat Quants & Generic APEX)
Bundled Q8_0 High-Precision Multimodal Vision Projector
Everyday Laptop Benchmarks (DDR4 / DDR5 RAM)
Estimated Projections on Consumer Hardware
The 24GB Miracle: Full 256K Context Runs In VRAM!
Hardware Throughput Projections (RTX 30 / 40 / 50)
Recommended Generation Parameters (XYZAILab Official)
Optional Support

[!NOTE]

ARCHITECTURE SELECTION GUIDE — MINIPLUS V2 & V2.1 EDITIONS

This repository hosts the MiniPlus V2 edition of XYZ-Aquila-mini. Our releases are precision-engineered for specific hardware budgets and memory topologies. V2 is NOT obsolete or "worse"; each edition serves distinct inference requirements:

MiniPlus V2 (High Theoretical Layer Protection): On paper, V2 provides extra protective envelopes on edge layers (10 layers in IQ3_S + IQ4_NL shared experts + Q8_0 attention gates). However, in practical inference benchmarks—even across extreme long-context windows exceeding +160K tokens—there is virtually NO perceptible difference in quality or reasoning compared to V2.1.

MiniPlus V2.1 (System RAM Streaming Specialist with Deep Context): Specially prepared to run totally or partially in system RAM (DDR4/DDR5) across long-context agentic reasoning windows (up to 256k tokens). By replacing non-linear codebooks with linear Q3_K edge experts, keeping Q8_0 attention gates, and upgrading shared foundation experts to Q5_K across all 40 layers, it completely eliminates AVX2 CPU dequantization stalls (+24 to 28+ tok/s streaming). Depending on your processor and memory bandwidth (DDR4/DDR5), streaming generation in system RAM can be almost as fast as having everything in VRAM, while supporting deep context keeping the dedicated Q8_0 multimodal vision projector (mmproj) explicitly loaded in GPU VRAM for real-time visual parsing. It provides this massive RAM streaming acceleration for only approx. 100 MB more, which is completely negligible in system RAM.

Which one should you choose? (Official Recommendation: V2.1)

⭐ PRIMARY RECOMMENDATION — XYZ-Aquila-mini APEX-I-MiniPlus V2.1: For virtually all users and deployments, V2.1 is the strictly recommended release. Empirically verified on WikiText-2, V2.1 achieves an outstanding Perplexity of 6.7927 ± 0.1884 (ΔPPL ≈ +0.07 from unquantized baseline (approx. 6.72)), matching the token fidelity of Q5_K / Q6_K class quantizations while weighing only approx. 14.7 GB (same footprint as Q3_K_M). Furthermore, it completely eliminates AVX2 CPU stalls, providing blistering +24 to 28+ tok/s streaming under system RAM offload.

MiniPlus V2 Legacy: Maintained for architectural transparency and users seeking specialized configurations for their workflow.

Both editions are handcrafted and vastly outperform flat 3-bit quants and generic community APEX-I-Mini releases. To explore or download the V2.1 edition of XYZ-Aquila-mini optimized for system RAM streaming, visit: IsValorum/XYZ-Aquila-mini-APEX-I-MiniPlus-V2.1-GGUF

[!WARNING]

DO NOT CONFUSE APEX-I-MINIPLUS WITH GENERIC COMMUNITY APEX-I-MINI!

Regardless of release version (whether V1, V2, or V2.1), NEVER confuse handcrafted APEX-I-MiniPlus builds with generic community APEX-I-Mini releases:

Generic Community APEX-I-Mini: Uniformly compresses all core MoE experts down to aggressive 2-bit IQ2_S (dropping below the critical quality floor), leaves the sensitive token output head unarmored at 3-bit Q3_K_M, and compresses attention projections down to Q3_K. In deep reasoning models, this triggers severe perplexity spikes, syntax errors, and broken code brackets.

Handcrafted APEX-I-MiniPlus (All Editions by IsValorum): Every single MiniPlus release—from V1 and V2 to V2.1—is a custom tensor-by-tensor architecture that preserves uncompressed F32 router gates, armors the token output head in high-precision Q6_K, safeguards attention gates in Q8_0, and keeps core reasoning experts at or above calibrated 3-bit (IQ3_XXS/IQ3_S). Even our earlier builds vastly outperform generic community APEX recipes and flat 3-bit quants.

[!TIP]

SYSTEM RAM INFERENCE: FULL OR PARTIAL

This APEX-I-MiniPlus release is designed for full or partial system-RAM inference. Depending on the processor, memory bandwidth, and DDR4/DDR5 configuration, generation can range from 20 to 45 tok/s. With partial GPU offload, systems that cannot fit 128K or more context entirely in VRAM can place the remaining model and context load in system RAM, maintaining stable, responsive generation at longer context lengths.

[!IMPORTANT]

EXPLORE THE ESTABLISHED 35B MoE MINIPLUS LINEUP

These are complementary APEX-I-MiniPlus V2.1 releases, not alternate downloads of the same model. Each receives the same tensor-by-tensor approach, integrated MTP where supported, and a design suitable for full or partial system-RAM inference. Choose the model whose native strengths best fit the work you want to do:

Qwen3.6-35B-A3B MTP APEX-I-MiniPlus-V2.1 — a versatile frontier MoE for broad reasoning, multilingual work, agents, tool use, and multimodal tasks.

Best for: General reasoning, agent workflows, tool calling, and flexible multimodal use.

Qwen3.6-35B-A3B MTP APEX-I-MiniPlus-V2.1 Abliterated — the V2.1 refusal-ablated Qwen3.6 edition for users who deliberately prefer reduced refusal behavior.

Best for: Workflows where an abliterated Qwen3.6 variant is explicitly desired.

Ornith 1.5 APEX-I-MiniPlus-V2.1 — a software-engineering-focused MoE designed for repository-scale coding and autonomous engineering agents.

Best for: Repository-scale development, multi-file code changes, and software-engineering agents.

Tiel Coder APEX-I-MiniPlus-V2.1 — a specialist coding MoE tuned for agentic programming, iterative tool use, and implementation-heavy work.

Best for: Focused coding sessions, iterative debugging, and tool-driven implementation.

These remain distinct model families and editions with their own behavior and empirical results. Pick by workload and intended alignment behavior rather than treating them as interchangeable quantization variants.

🏅 Independent Benchmark of the APEX-I-MiniPlus Family (Occamy V2 Reference)

[!NOTE] External report: zephel01 independently benchmarked Occamy V2. The benchmark below was performed on Occamy-1.0 APEX-I-MiniPlus V2, not on this specific model. It is included as independent evidence of the broader MiniPlus quantization approach.

The APEX-I-MiniPlus quantization architecture powering this model was subjected to an extensive independent evaluation by Japanese AI researcher and evaluator zephel01 (CoolZero) on an NVIDIA RTX 5090 (32GB) workstation running llama.cpp CUDA b11027 with FlashAttention (-fa on -ctk q8_0 -ctv q8_0 -ngl 99).

The evaluation tested the APEX-I hybrid MoE engine across 348 unseeded trials on SWE-bench style multi-file Python bug-fixing tasks with hidden pytest suites (llmbench):

L6 Multi-File Code Generation (60 tasks):
- Context 32,768 (32K): 93.3% Resolved (46/60 tasks passed 5/5 consecutive trials; 20/20 on Easy–Hard).
- Context 65,536 (65K): 90.0% Resolved (45/60 tasks passed 5/5 consecutive trials).
- Match with 25–28 GB Models: Matches or exceeds the resolution rate of full 25–28 GB models (such as Ornith-1.5 and Tiel-Coder 35B-A3B) while consuming over 10 GB less VRAM (14.6 GB vs approx. 26 GB).
Extreme Context VRAM Scaling (The Hybrid DeltaNet SSM Advantage):
- 32K Context: 14.6 GB total VRAM allocation.
- 65K Context: 15.1 GB total VRAM allocation (only +0.5 GB VRAM added when doubling context!).
- Architectural Explanation: Because 30 of the 40 layers utilize Linear Attention / DeltaNet SSM ($O(1)$ constant recurrence memory), only the 10 full-attention anchor layers expand the KV cache. This proves empirically that 65,536 context runs 100% in VRAM on consumer 16GB GPUs (RTX 4080 / RTX 5080) without offloading to system RAM.
Measured Real-World Throughput: Sustained single-stream generation of approx. 247 – 251 tok/s on NVIDIA RTX 5090.

Bundled Model Files & Specifications

File Name	File Size	Memory Footprint	Format / Precision	Purpose
`XYZ-Aquila-mini.APEX-I-MiniPlus-V2.gguf`	`14.63 GB` (`13.63 GiB`)	`13.63 GiB`	Custom APEX-I (3.38 BPW)	Main agentic search, browser reasoning & logic core
`mmproj-XYZAILab_XYZ-Aquila-mini-Q8_0.gguf`	`610 MB` (`582 MiB`)	`582 MiB`	High-Precision `Q8_0` Projector	Required for browser viewport inspection, UI clicks & OCR

Base Architecture: Qwen3_5MoeForConditionalGeneration (40 layers, 256 fine-grained micro-experts with intermediate dimension 512, 8 active per token) + Vision Projector.
Active Parameters: approx. 3.2B active parameters per token (high-throughput streaming paired with 35B multimodal depth).
Importance Matrix: Calibrated on dense multimodal search traces, browser DOM interactions, and visual question answering datasets.
Memory Footprint: Lean 13.63 GiB weight footprint engineered to avoid OOM crashes on 16GB and 24GB VRAM setups.

Comparative Quantization Analysis (vs. Flat Quants & Generic APEX)

Also, don't confuse APEX-I-MiniPlus-V2 with a generic baseline APEX-I-Mini. Traditional APEX-I-Mini drops core experts aggressively to 2-bit IQ2_S and leaves output.weight at 3-bit Q3_K_M, which creates a noticeable perplexity hit on complex reasoning tasks. V2 was specifically re-engineered to avoid that quality floor (keeping core experts at calibrated IQ3_XXS, output in Q6_K, shared expert in non-linear IQ4_NL, and routers in F32).

To put the numbers in perspective: this cuts nearly 2 GB off a flat 3-bit quant (approx. 15.6 GB), and weighs only about approx. 1 GB more than a generic APEX-I-Mini (approx. 12.5 GB). For that single extra gigabyte of VRAM, you get a massive jump in reasoning and syntactic stability.

Take a look at the tensor-by-tensor comparison table below to inspect the exact architectural differences and see why this specific allocation is optimal. That's specifically what this was built for:

Architectural Component	Generic Automated Quants (Flat `Q3_K_S` / `IQ3_S`)	Generic APEX-I-Mini (Baseline Recipe)	Our Handcrafted APEX-I-MiniPlus-V2 (IsValorum)	Perceived Quality & Real-World Impact
Output Head (`output.weight`)	Flat `IQ3_S` / `Q3_K_S` (approx. 3.44 BPW)	Inherits base type `Q3_K_M` (approx. 3.44 BPW unarmored)	`Q6_K` (approx. 6.56 BPW uncompromised)	Eliminates Syntax & Vocabulary Hallucinations: Low-bit output heads cause tokenizer classification noise, breaking code indentation, brackets (`{}`, `[]`), math symbols, and domain terms. `Q6_K` preserves near-FP16 output classification.
Expert Routers (`ffn_gate_inp.weight`)	Blindly quantized to 3-bit / unoptimized	Inherits base type `Q3_K_M` (approx. 3.44 BPW compressed)	`F32` uncompressed (32.0 BPW, 2 MB/layer)	Zero Router Drift: In micro-expert models, even minuscule quantization errors in router logits misdirect tokens to wrong experts. Retaining uncompressed `F32` guarantees 100% routing fidelity with virtually zero memory overhead (approx. 80 MB total).
Attention & Language (`attn_output`, `attn_qkv`)	Flat `IQ3_S` / `Q3_K_S`	`Q3_K` on 34 middle layers (L3–36), `Q4_K` on 6 edge layers	`Q6_K` for `attn_output`, `IQ3_S` for `attn_qkv`	Contextual Retrieval Precision: Generic APEX reduces attention and language projections to `Q3_K` across 85% of layers. Our V2 build protects attention output in high-precision `Q6_K` and uses calibrated non-linear `IQ3_S`, ensuring flawless needle-in-a-haystack retrieval across deep 128k–256k context windows.
Attention Gates (`attn_gate.weight`)	Blindly compressed to 3-bit	Compressed to `Q3_K` (middle) / `Q4_K` (edges)	`Q8_0` (8.50 BPW)	Attention Head Stability: Attention gates modulate query-key routing across hybrid attention layers. Keeping them in 8-bit prevents attention crosstalk and hallucination over long contexts.
*Shared Foundation Expert (`ffn__shexp`)**	Flat `IQ3_S` / `Q3_K_S` (3.44 BPW)	Linear `Q4_K` (middle) / `Q5_K` (edges)	`IQ4_NL` (4.50 BPW non-linear codebook)	Foundational Knowledge Armor: The shared expert executes for 100% of tokens. In 256 micro-expert models, `IQ4_NL` non-linear codebooks preserve heavy-tailed outlier representations far better than standard linear quantization.
Core MoE Layers (Middle: 10–29)	Flat `IQ3_S` / `Q3_K_S` (uniform bit-rate across all layers)	Aggressive `IQ2_S` (2.50 BPW)	`IQ3_XXS` (3.06 BPW) + calibrated `imatrix`	Above the Quality Threshold: Generic 2-bit `IQ2_S` baselines drop below the critical quality floor for 35B MoEs, resulting in perplexity spikes on reasoning tasks. Our `IQ3_XXS` with imatrix achieves deep compression (272 MiB → 98 MiB per block) without sacrificing logic.
Edge MoE Layers (Layers 0–9 & 30–39)	Flat `IQ3_S` / `Q3_K_S` (no layer-wise gradient)	`Q3_K` (limited to first/last 5 layers only: L0–4, L35–39)	`IQ3_S` (expanded to 10 input & 10 output layers)	Protected Ingestion & Synthesis: Half of the model's layers (10 at input, 10 at output) form a non-linear armored envelope, preventing prompt misunderstanding and token degeneration across 256 micro-experts.
Multimodal Vision (`mmproj`)	Often omitted, or left as uncompressed `FP16` (approx. 900 MB)	Often omitted or separate uncompressed `FP16`	Bundled `Q8_0` (582 MB) with 27 critical F32/F16 fallbacks	Saves approx. 320 MB VRAM with Zero Loss: Handcrafted quantization preserves normalization and bias tensors in F32/F16, ensuring razor-sharp OCR, DOM viewport reading, and coordinate detection without visual noise.
Normalization & Biases	Often degraded	Standard	`F32` uncompressed	Numerical Stability: Prevents cumulative floating-point underflow/overflow across deep 40-layer computation.

Bundled Q8_0 High-Precision Multimodal Vision Projector

Standard community quants frequently omit the multimodal projector or supply uncompressed FP16 files (approx. 857 MB), doubling visual memory overhead.

Bundled Q8_0 Projector: Pre-quantized to Q8_0 (582 MiB / 610 MB), saving approx. 300 MB of VRAM.
Audited Layer Fallbacks: llama.cpp automatically preserved 27 critical normalization and bias tensors in F32/F16, ensuring razor-sharp rendering of browser DOM text, minute UI action targets, and dense infographic diagrams.

Everyday Laptop Benchmarks (DDR4 / DDR5 RAM)

Estimated Projections on Consumer Hardware

You do not need an enterprise server to run an autonomous multimodal web agent. Estimated throughput projections on an everyday consumer laptop (Intel Core i5 / AMD Ryzen, 4GB/6GB Laptop GPU, 32GB DDR4/DDR5 RAM):

GPU VRAM Allocation: Uses only approx. 3.8 GB VRAM (fits effortlessly on budget laptop GPUs like RTX 3050, 4050, or 2060).
System Memory Offload: Standard 32GB system RAM accommodates the remaining layers.
Estimated Document / Screenshot Ingestion (Prefill): 300 to 420+ tokens/second sustained across full viewport inputs.
Estimated Streaming Generation: 20 to 24+ tokens/second sustained output across system RAM!

The 24GB Miracle: Full 256K Context Runs In VRAM!

Autonomous web search and UI navigation rapidly fill context buffers with full-page DOM trees, image embeddings, and multi-turn action traces. Generic community quants weigh 16–19 GiB in weights alone, immediately crashing 24GB cards.

XYZ-Aquila-mini APEX-I-MiniPlus-V2 fits the entire 256K context window within 24GB VRAM:

Context Length	Model Weights (Est.)	KV Cache (q8_0, 4 slots)	Compute Buffers	Total GPU VRAM (Est.)	Hardware Feasibility
32,768 (32k)	`13.63 GiB`	`0.58 GiB`	`1.80 GiB`	`16.01 GiB`	Full offload on 24GB; partial on 16GB
65,536 (64k)	`13.63 GiB`	`0.92 GiB`	`1.95 GiB`	`16.50 GiB`	Effortless fit on 24GB GPUs
131,072 (128k)	`13.63 GiB`	`1.58 GiB`	`2.22 GiB`	`17.43 GiB`	Effortless fit on 24GB GPUs
262,144 (256k)	`13.63 GiB`	`2.92 GiB`	`2.80 GiB`	`19.35 GiB`	FULL 256K AGENT TRACE IN VRAM!

Note: Projections leave approx. 4.65 GiB of headroom on 24GB cards for display buffers and the Q8 vision projector.

Hardware Throughput Projections (RTX 30 / 40 / 50)

⚙️ Recommended Generation Parameters (XYZAILab Official)

Official generation parameters specified by XYZAILab for structured synthesis, tool use, and multilingual reasoning:

Hyperparameter	Value	Description / Creator Configuration
Temperature	`1.00`	Official XYZAILab setting for deep reasoning and multi-turn tool calling.
Top-P	`0.95`	Calibrated nucleus sampling threshold.
Repetition Penalty	`1.05`	Prevents token cycling in long reasoning trajectories.
Max Tokens	`32768`	Official token budget for complex structured outputs.
Thinking Flags	`true`	Official config: `enable_thinking: true, preserve_thinking: true`.

[!IMPORTANT]

🔍 Model Inherent Behavior vs. Quantization Fidelity Notice

Any behavioral nuances, stylistic tendencies, domain-specific habits, or zero-shot edge-case oversights stem entirely from the original unquantized checkpoint weights and fine-tuning distribution, NOT from the APEX-I quantization process. Handcrafted APEX-I-MiniPlus strictly preserves mathematical tensor fidelity—keeping 100% of expert routing matrices (gate_inp) in uncompressed F32 (zero router drift), armoring the token output head in Q6_K, and safeguarding attention gates in Q8_0. Empirical verification confirms near-zero perplexity loss (ΔPPL ≈ +0.07), ensuring that token logits, routing decisions, and reasoning trajectories are mathematically faithful to the original base model.

7. Community Compute Fund & Student Support

All IsValorum quantizations will always remain completely free and open to the public without paywalls.

However, cloud GPU compute is expensive and funded entirely out-of-pocket. If you find these handcrafted builds to be of high quality and they benefit your day-to-day workflow, you can support future Day-0 releases on Ko-fi. Every contribution directly funds the cloud compute hours required for new model runs and helps support my university education:

(When supporting on Ko-fi, feel free to leave a note with your Hugging Face handle and the specific model you would like prioritized).